enges of imp act assessm ent - · pdf fileacademic and non-academic impact of arts and...
TRANSCRIPT
The
Work
ESF Me
eChall
king Grou
ember Organ
enges
p 2: Impa
nisation Foru
ofImp
ct Assessm
um on Evalua
1
pactAs
ment
ation of Pub
ssessm
licly Funded
ment
Research
2
Table of contents
1. Introduction
Background
The aim of the report
The structure of the report
2. What do we mean by impact?
Outputs, outcomes, impacts
Instrumental, conceptual and broad consequences
The nature of impact
Conclusion
3. Impact studies – methods and methodological challenges
Attribution
The counterfactual argument
Time lags
Micro/macro level
4. Impact study methods – as discussed in the literature
Input measures
Output measures
Expert reviews
Surveys
Case studies
Hindsight studies
Economic Models
Questions asked and appropriate methods
5. Recommendations
References
List of WG members and Acknowledgements
3
1. Introduction Background
What is the purpose of and how can one assess impact from research? This is an important question
as societies are dependent, and increasingly so, on the creation of knowledge to yield innovations
and better policies to generate economic and social benefits. The assessment of impact from
research is the attempt to establish to what degree research affects certain changes in science and
society. What evaluation means for research funding organisations is discussed in a report from an
earlier European Science Foundation (ESF) Member Organisation Forum (European Science
Foundation 2009). This study identified five different types of evaluations, and considered different
levels of evaluations: a) of the funding agency as an organisation; b) of funding policies; c) of research
fields or scientific disciplines; d) of funding schemes; and e) of research grants.
There is not always a clear-cut distinction between impact assessments and other evaluation studies.
In general evaluations of research the prime objective is to support and strengthen the quality of
research, but impact assessments have a broader objective. Impact assessments can share some
characteristics with a goal oriented type of evaluation. While the latter aims to evaluate how the
goals and objectives of the intervention (for example, a research project) are being met, the impacts
do not necessarily have to be formulated as goals for the intervention. One of the aims with impact
assessment is to gain a more thorough knowledge of the relationship between actions and effects,
but this can also be an aim for formative or process oriented evaluations. It is therefore important to
make a distinction based on the purpose of the evaluation.
Ideally, the use of impact assessments could perform a dual task: demonstrating the value of
research, and increasing the value of research through a more effective way of financing research in
order for research to have impact. Firstly, publicly funded research has a responsibility to contribute
something in return to society, and impact studies are one way of showing these returns and, in
doing so, legitimise investments in research and serve as instruments to advocate for funding.
Secondly, impact studies can strengthen returns to science and society by improving the instruments
that are used to fund research. In this instance they may also provide a better understanding of
transfer of scientific knowledge into practice. The improvements in instruments may be structural,
i.e., improving the way funding schemes are constructed and how research environments interact
with society, or improvements may also arise as the process of evaluation affects the individual
behaviour of researchers and stakeholders, as they become more aware of how research affects
society and vice versa. They can also be used to further discuss the question of the relationship
between scientific excellence and research being beneficial to society.
The interest in understanding the links between research, innovation and societal development is not
new and neither is the search for appropriate methods to assess the impact of research. In the last
few years, however, there has been growing pressure from the political sphere to measure and
achieve value for money from investments in research. The complex and interrelated reasons behind
this recent development is not the focus of this report but a discussion of impact studies cannot
disregard the relationship between the STI (science, technology and innovation) policy framework
and an evaluative framework (Donovan 2007a and 2007b). It is argued that “research policy has been
broadly reframed emphasising notions of ‘value for money’, democratic oversight and
accountability” and that “[p]ublic research funding is … increasingly understood as a strategic
4
investment where state economic and regulatory strategies are oriented towards maximising
returns” (Kearnes and Wienroth 2011, p. 157). In this context, public research funders are assigned a
new role where they are players that transcend funding research, and research and the benefits of
research are more closely linked to other sectors of society.
Most notable is the increasing pressure to demonstrate impact in basic research. The debate on the
economic value of public research funding has therefore invoked “… a set of contested cultural
values concerning the role of science in contemporary social life” and can be seen as “… a historic
contestation over the standards by which the principles of creativity, autonomy, and diversity are to
be judged” (Kearnes and Wienroth 2011, p. 155). The research sector has reacted by emphasising the
long-term economic and social values associated with basic research and arguing for the importance
of assessment of the quality of knowledge produced. That this policy shift has caused not only
debate but also some anxiety is understandable as much might be at stake if assessments of impact
become a predominant deciding factor in the distribution of public research funding (see, for
example, the REF Pilot a 2009; b 2010; c 2010).
This raises the question if it is at all feasible to assess impact of all kinds of (basic) research. A NIST
report from 2003 points to the general methodological problems associated with impact assessment
(Tassy 2003, p. 1):
In fact, a single ‘manual, for impact assessment may never be achieved. The technology
trajectories and economic outcomes that government programs and projects seek to leverage
vary significantly, as do the complex economic structures that characterize a technology-based
economy. Thus, no metric or measurement method can (1) address the diversity and complexity
of an R&D agency’s technological outputs, (2) describe the subsequent processes by which
private sector impacts occur, and finally (3) accurately capture the resulting economic outcomes.
Resulting from an awareness of the shortcomings of the instruments for assessment of impact, a
significant amount of energy has been directed towards the improvement of established methods
and towards the development of new methods of assessing impacts. Special attention has also been
paid to developing methods that aim to capture values other than economic outcomes. This
development and the problems as well as advantages associated with different methodologies will
be discussed in more detail in this report.
The aim of the report
For the member organisations of ESF, how to conduct ex-post impact evaluations is a burning issue.
Demonstrating and achieving returns to society from research are high on the agenda. The prime
instrument for this task – impact assessment – is difficult.
This report aims to highlight the development of different methodologies for assessing impacts. We
will discuss the advantages and shortcomings of different methodologies. The focus is on impact
assessment in relation to public research funding, and is especially concerned with the evaluation of
funding schemes and research programmes. The main concern is to identify good practices of impact
evaluations and to make recommendations on how to perform impact evaluations.
The structure of the report
5
We will begin by discussing the concept of impact; what does impact mean? The main challenges
with impact evaluations will then be highlighted and further discussed. We will then proceed to
discuss the various approaches to assessing impact that exist and we will highlight areas of special
interest to the member organisations of the ESF. The report concludes with recommendations for
the future.
2. What do we mean by impact?
In the literature there are different conceptions about what impact is, but it is surprisingly often left
without a definition. Impact can be described as consequences of an action that affects people’s lives
in areas that matter to them. The consequences would not have occurred without the original action.
It can be added that not all consequences may warrant an assessment of impact. They must be of
some importance to us, or seen as useful.
Impact does not necessarily have to beneficial. There can also be different perspectives as to
whether a particular impact is positive or negative. Impact does not have to be intended. Nor should
it be seen as final consequences of an action, as we can never know when the consequences of an
action are final.
At an overarching level it is possible to make a distinction between scientific consequences and
societal consequences of research. Scientific consequences are, for example, the advancement of
knowledge and how the research landscape is influenced. For example, what consequences does the
funded research have on the organisation of research (size, cooperation, human capital)? If there is
investment in a certain field of research, how does this affect this field scientifically and how are
related fields affected? And with regards to research questions within this field: To what extent are
the research questions directed towards the needs of society? To what extent are the research
questions risky?
Societal consequences include addressing questions, such as what does society gain in the form of
better products, better services, healthier lives, better welfare, a sustainable development, etc?
Hence, impact on society can include every important aspect of society. Early studies on assessing
impact concentrated on what science contributed to technological innovations, improvements in
health and the creation of national wealth. Yet these benefits are still most commonly associated
with what science is supposed to deliver to society. There has been a considerable broadening of
impact assessments to include, for example, cultural and social returns to society, although studies of
academic and non-academic impact of arts and humanities, conceptually and methodologically, are
still underdeveloped compared to other subjects (Levitt et al. 2010, p. 35). However, in the above
examples of impact there is a difference in how tangible they are. Whereas some are very noticeable
– economic wealth – some are less easy to measure, for example, changing attitudes and ways of life.
Nevertheless, not all important consequences of an action need to be labelled impact, which will be
discussed below. But, the increasing interest in the less tangible impacts have changed the methods
used to assess impacts; a move from a heavy reliance on quantitative measures, often on a macro
level, to a wider set of methods of used.
6
Outputs, outcomes, impacts
Sometimes, there is a distinction drawn between outputs, outcomes and impacts of research, all of
which are consequences of an action. Publications, discoveries, patents are normally seen as outputs.
These outputs can become outcomes: for example, policy guidelines, building competence and
product development. In some research impact assessments there are other terms besides outputs
and outcomes, but the logic remains consistent. Impact is then described as increased employment,
improved health and/or wellbeing and increased productivity and reduced waste.
Sometimes a conceptual framework of outputs/outcomes/impacts is portrayed as a linear process.
According to the linear model the research process is a one-way street. Knowledge from basic
research transforms into applied research questions and finally into social and technical innovations,
and then impact. A linear model of how research translates to impacts has been criticised for not
corresponding well with reality. Impact sometimes also relates to final outcome, which is
problematic as it is impossible to know when the outcome is final. Instead, in the conceptual
framework above, it would be better to describe impact as a broad effect on science or society.
But the conceptual framework does not have to be portrayed in a linear fashion. The
output/outcome/impact perception may be useful to give a structure to the analysis without
presupposing a linear development from research to practice. However, the dividing line between
outputs, outcomes and impact is not always that clear-cut.
Instrumental, conceptual and broad consequences
Another way to conceptualise impact is to see impact as something all-encompassing rather than as
in the above distinction between outputs/outcomes/impacts. Consequences could instead be
described as instrumental, conceptual or broad. The idea behind this theoretical model is that the
interaction between research and society affects how different actors in society behave, either by
direct influence or by influencing their opportunities to act. This could be applied in numerous ways,
for example, through policy – a plan of action or a measure developed in response to a perceived
need, in order to achieve a particular outcome; through public debate – in what aspects does science
affect the public debate – topics, how the public debate is structured; or through medical treatments
and new technical appliances.
Consequences of an action on a societal level could then be seen as the following examples (see
Davies, Nutley and Walter 2005):
Instrumental consequences: political decisions, official guidelines, new technology, spin-off
companies
Conceptual consequences: knowledge, understanding and attitudes in certain issues
Broad consequences: economic welfare, improvements in health, sustainable environment.
And consequences on science can be seen as:
Instrumental consequences: publications, organisation of science, collaborations, training of
personnel
Conceptual consequences: ways of thinking, new research questions
7
Broad consequences: paradigm shifting knowledge.
Which of these consequences should be seen as impact? The question is not straightforward to
answer, and it is difficult to give an authoritative answer. If transformed to the logic behind the
outputs/outcomes/impact model at least the broad consequences are to be regarded as impact. But
the other consequences or parts of them may also be seen as impacts.
The nature of impact
As already mentioned impact can roughly be divided into scientific and societal. Societal impact in
turn can be divided into several categories (see, for example, European Commission 2010; Delanghe
and Teirlinck 2010).
Scientific impact: contribution to the subsequent progress of knowledge, the formation of
disciplines, training and capacity building.
Technological impact: contribution to the creation of product, process and service
innovations.
Economic impact: contribution to the sale price of products, a firm’s costs and revenues
(micro level), and economic returns either through economic growth or productivity growth
(macro level).
Social impact: contribution to community welfare, quality of life, behaviour, practices and
activities of people and groups.
Political impact: contribution to how policy makers act and how policies are constructed and
to political stability.
Environmental impact: contribution to the management of the environment, for example,
natural resources, environmental pollution, climate and meteorology.
Health impact: contribution to public health, life expectancy, prevention of illnesses and
quality of life.
Cultural impact: contribution to understanding of ideas and reality, values and beliefs.
Training impacts: contribution to curricula, pedagogical tools, qualifications.
All of these forms of impact may be interesting to investigate, but what forms of impact that are
assessed do also affect what methods that are the most appropriate. Different methods and
approaches to impact assessment will be discussed further in chapters 4 and 5.
Conclusion
The point of making all these distinctions is not to recommend that every one of these distinctions
needs to be applied. But, when designing an impact assessment, it is advisable to conceptualise what
impact is to be assessed in order to make a study that is fit for purpose.
To summarise, impacts are consequences of importance of an action, and the focus of impact
assessments can be both impact of research and impact on society broadly defined. Before a study is
undertaken it is useful to clarify the concept of the impact, and if and how it differs from other
consequences of an action.
8
3. Impact studies – methods and methodological challenges
To assess impact of research poses serious methodological challenges and there are at least four
crucial problems that have to be taken into consideration:
1) How to attribute the intervention to the observed effects.
2) How to determine counterfactual positions, i.e., would the observed effects have occurred
anyway?
3) How to deal with the time lags between research and tangible outcomes, and the multiple
stages inbetween.
4) Where to focus the assessment, ranging from a research project to research in general.
The answers to these questions have a profound effect on the design and methodology of an impact
study.
Attribution
The general idea behind attribution is to link impact to a certain instrument or action (e.g., a research
programme). If it can be established that an observed phenomena (e.g., economic growth) is more or
less dependent on the instrument or action, we can state that the observation is attributed to the
action taken. However, the means of convincingly establishing an attribution may be insufficient,
and some paths to impact are more obscure than others. To isolate the role played by a single actor
or action, such as a research funder, is consequently cumbersome or more likely virtually impossible.
Thus, from a methodological point of view, it is very difficult to establish attribution. It is also more
difficult if the object of analysis is on a macro level, i.e., overall economic impact, general improved
health, etc., as there will be a multitude of factors affecting the outcome. Some impacts appear many
years, even decades, after the research is carried out. This makes it even more difficult to deal with
attribution.
The recommendation is to focus on contribution or the value of the research (Levitt et al. 2010,
p. xiii), rather than attribution, of a funder, programme or project, i.e., to say that a funder has
played a role in the impacts of a research project it has funded, rather than determining the exact
share of the impact that can be claimed.
There are two reasons to abandon attribution for contribution. Firstly, we know that the outcomes of
research are very rarely the consequence of one singular research activity, and whether research
transforms into impact is also dependent on factors and actors outside the research sector.
Consequently, it is a false way to describe reality. Secondly, from a methodological point of view it is
very difficult to attribute outcomes to inputs and it demands a number of assumptions (of which
some surely will be over-simplified) (Cox 2010, pp. 10-15).
Research funders who wish to establish what difference their specific funding makes might still be
interested in attribution, and rightly so. Attribution, rather than contribution, is also probably a
better basis for making future strategic choices. But the urge to establish attribution should not stifle
the possibilities of doing impact studies. So to raise attribution as a real obstacle for doing impact
9
assessments is in some way misdirected. To strive for attribution is understandable, but the
possibilities of the methods that can be used are limited. To conclude, there is not a clear-cut
distinction between attribution and contribution but more of a sliding scale. As far as possible one
should try to establish the magnitude of the influence of an action. But, this should not be done at
the expense of the rigour of the methodology.
The counterfactual argument
Related to the question of attribution is the challenge of defining an appropriate counterfactual
position. To determine the impact of an intervention, one must also estimate what would have
happened if the intervention had not taken place, as it may be possible that the outcomes we are
tracing might have occurred anyway. Even more difficult is the question of opportunity costs. Would
the same resources have provided more impact if they had been invested in another research project
or programme (Go8 2011, pp. 6-7)?
To establish a counterfactual position is relatively easily done in a laboratory setting where essential
factors influencing the outcome can be held constant. Another way to establish a counterfactual
position is to use a randomised controlled trial (RCT), a method often used in medical science. The
key feature of a typical RCT is that the study subjects are randomly allocated to different groups,
which are distinguished by how they are affected by the intervention or not. Hence, the results will
show one (or more) group that is affected by the intervention and another group that is not affected.
The differences that can be observed between the two groups can be described as the impact of the
intervention.
Generally, methodological discussions on how to establish a counterfactual position are somewhat
underdeveloped in impact assessments of research. Ramberg and Knall (2012) do extensively discuss
the challenge of establishing a counterfactual position, but from a theoretical perspective. There are
very few examples of elaborating with counterfactual positions in impact assessments, and when
assessing impact of research it is practically impossible to establish a counterfactual position based
on experiments or quasi-experiments.
What other measures are then appropriate to establish a counterfactual position? There are various
techniques available, which it is advisable to combine, to establish a satisfactory counterfactual
position.
1) Interviews
One way is to interview researchers who have been financed or in other ways affected by the
action or stakeholders who are likely to be affected by the action. They are in a position to
judge what difference the action has made. In this way, it is possible to compare those have
been exposed to an intervention with the situation before the exposure. However, this is also
a source that has its weaknesses. Those who are interviewed may have difficulties
remembering and estimating the impacts in an impartial manner.
2) Establish a model of the intervention logic
Intervention logic models are often used in impact assessments. In such a model the search
for impacts follows a description of how impact is created in an ideal situation. It involves a
description of how impacts are thought to materialise. If impacts can be observed that fit
10
into the intervention logic there is a good indication that the action has caused the impacts.
A useful model must strike a balance between usefulness and realism. It cannot portray
reality fully as that would obscure its purpose. At the same time it needs an adequate
realism, otherwise it would be unlikely to produce interesting results. An intervention logic
model needs to include expected outputs (or consequences) from research, and an idea of
how diffusion of knowledge to society takes place and how actors can make use of the
research, and finally the connection between these levels (Molas-Gallart, Tang and Morrow
2000, p. 174)
A much used intervention logical model is the payback framework (see Hanney et al. 2004),
but there are also other examples (see Academy of Finland 2009, pp. 21–24; Molas-Gallart et
al. 2000, p. 173). Using intervention logic gives the possibility to investigate if expected
impacts do not materialise. And it makes it possible to distinguish if the lack of impact is
caused by a flawed logic in how it is thought to be materialised or if there is a problem in the
phase of implementation. There is really no contradiction between an intervention logic
model and capturing impacts outside the expected logic of impact. In the design of the
impact assessment it is advisable, depending on the aim of the study, to include elements
where unexpected impacts can be captured.
3) Judgments by experts
Identify different actions (inputs), including others that are of primary interest to the impact
assessment, and consequences (that may translate into impacts). Can impact be established
or are the consequences caused by something else? In this process it is possible to have
experts to make judgments on what the situation would have been if the intervention had
not occurred.
4) Establish control groups
If possible, try to use some form of control group. One way of constructing a control group
could be to compare those who have been exposed to an intervention to the typical
outcome in a group that has not been exposed to an intervention. This is almost a quasi-
experimental design, but not as strict and controlled. It could, for example, be a comparison
between researchers who were funded in a certain programme and those who were
rejected. There are very few examples of some form of control group being used in impact
studies, and it would be interesting to see a wider use of this technique.
Time lags
All studies acknowledge that the contributions of research often occur and manifest themselves over
long timescales. The short-term impact of research can, furthermore, differ significantly from the
long-term impact from the same research (Go8 2011, p. 4).
When is then a good time to measure impact? For how long should we be trying to identify impacts?
There is, of course, no authoritative answer to these questions. On the one hand, there is a need for
a sufficiently historic time window to allow impacts to occur. It might take a very long time before
broader impacts from research appear, not least from basic research. In many cases it is not possible
11
or even desirable to wait that long to conduct an assessment of impact. As impact evaluations are
often intended to serve as a basis for strategic planning, the information may not be useful after a
long period and is required immediately. On the other hand, collecting evidence is necessary, and
this must be done without losing too much detailed information and data. What is the quality of
records, the ability of researchers to recall their activities and stakeholders to remember how
practice was influenced in a process that has proceeded for a number of years?
Ideally, an impact study with such a profile is best suited to be conducted alongside this process. But
impact studies running for 10-20 years are not very likely, and there would be very few, if any,
executed. Furthermore, in many cases, the question most likely to attract interest is what recent
investments in research have led to.
How can this situation be dealt with? First of all, there is a need for studies that can be described as
longitudinal, and which create a foundation for the future understanding of impact. These studies
follow a long time frame and try to capture impacts in the long run as well as in the short run. This
kind of study also describes the process from research to impact.
These longitudinal studies have, apart from their own aims, a certain purpose for impact evaluations
that are not longitudinal. If you lay out the full process from research to impact you will hopefully
notice that there are short-term indicators that are predictive of long-term impact. These early
success indicators presuppose that you have detailed information about what causes impact.
The recommendation is to do more research and studies into the processes of how impact occurs.
This would hopefully lead to good and solid models and theories of how research gives impact. And,
even more hopefully, this could pave the way for finding short-term indicators that are predictive of
long-term impact. This would in addition lessen the problem of the time lag, and thereby impact
studies may play an even larger part in influencing the strategy of research funding. There is
definitely some interesting work being done in this area, for example the SIAMPI model that will be
elaborated in the next chapter (see also Canadian Academy of Health Sciences 2009; Luoma et al.
2011 for an analysis of indicators).
Another similar way to deal with the time lag is to see the impact in different phases, which will leave
room for different approaches to analysing impacts:
a) Potential impact (short-term), for example research on knee surgery that has the potential to
be more effective so that people can be rehabilitated faster and use less health care
resources.
b) Action towards impact (medium-term), new guidelines on how to perform knee surgery.
c) Impact observed (long-term), operations are carried out according to the new guidelines
which lead to benefits for the individual and society.
Micro/macro level
Another challenge in impact assessments is the object of analysis and what conclusions and
recommendations that is possible form an impact assessment, and that depends a lot on what level
the impact assessment is carried out.
12
A study from a macro perspective tries to answer the impact of research in general and the impacts
of research are on a high level of aggregation, for example, economic wealth. A micro perspective on
the other hand takes a specific research project as the point of departure. Of course, there are
several approaches between the two extreme positions described above.
The obvious drawback with a macro approach is the difficulty of determining if the consequences
observed really are impacts of the action assessed. A correlation is easier to establish with a micro
approach, but the drawback of this approach lies in the difficulty of aggregating the results from a
micro approach to a general level.
4. Impact study methods – as discussed in the
literature
Impact assessment methods and methodological challenges will be further discussed in relation to
the reviews of the selected accomplished impact studies. As an introduction to these reviews we will
in this part present some methodological approaches that have been discussed in studies of impact
assessment. It is possible to categorise the methods or models in several ways, for example, on the
basis of the type of impact, the level of aggregation (national research funding or specific research
programmes), the expected audience for the results, whether the assessment is quantitative or
qualitative, etc. Some methods can furthermore be used for different types of impact studies and for
different purposes, and many of the more complex models presuppose that various methods are
used in combination. This is also the case with most impact assessments. The aim here is to give a
brief overview and the different methodological approaches are presented according to a simple
grouping, which includes a mix of methods and what could be called impact assessment models.1
Input measures
Input measures do not identify impact but they can give vital information about the kinds of impact
one might expect. On a national level the information about the prioritising between basic and
applied research, between different research fields, etc., as well as information about the balance
between different sources of research funding can serve as indicators. To understand the boundaries
for possible impacts in this way can be an important and relevant starting point also for individual
research funders.
Output measures
1 For this part a combination of several articles and reports has been used and they are all included in the list of
references. Direct references will be given only when we are referring to specific information.
13
Output measures can be used to assess research productivity, and they can be seen as a stage in the
road from research to impact. The most obvious output measure is publications, but processes or
tools used to disseminate research can also be considered as a type of output indicator. Both types
will be discussed in the following section.
Bibliometrics
Today publications and especially citations tend to be looked upon as indicators of impact rather
than merely output measures. Bibliometrics has experienced a rapid and immense development and
the methods have become more sophisticated and more commonly used. The use of bibliometrics
has, however, also been subjected to severe critique. The critique has highlighted several
problematic issues. One shortcoming with these quantitative indicators is that they do not grasp the
qualitative aspect of research excellence and although citations say something about the impact one
must remember that the citations can be both positive and negative. Another problem is associated
with the differences in publishing cultures that disadvantage some disciplines compared to others. In
order to overcome this, field sensitive citation indexes have been constructed. Not all are convinced
that this will solve the whole problem and in a critical essay on research evaluation, Claire Donovan
(2007b, pp. 591-592) raises the question if the novel metrics should not be looked upon as
palliatives. Still, bibliometrics seems to serve a certain function in impact assessment, at least for
assessing research impact. When it comes to impact in a broader sense, other types of output
measures should be used as indicators, which we will discuss next.
Dissemination of research and interaction between science and society
Different ways of disseminating research can be seen as output measures in that they can potentially
lead to impact. “These can encompass the use of technology transfer mechanisms such as industry
seminars, industry secondments … field days … participation in government committees and policy
development processes, participation in industry and academy meetings and seminars, preparing
popular publications, research consultancy work …” (Go8 2011, p. 12). In the reports from the impact
pilot exercise for the REF 2014 in UK these kinds of indicators are presented as possible instruments
that the universities can make use of for assessing impact - defined broadly to include social,
economic, cultural, environmental, health and quality of life benefits (REF Pilot a 2009; b 2010; c
2010). Another UK example is discussed in an article by Matthew Kearnes and Matthias Wienroth on
the Engineering and Physical Sciences Research Council (EPSRC) and its way of responding to the
contemporary policy discourses concerning the impacts of public research funding. In order to
supplement the methods of assessing research impact the EPSRC has introduced the notion of
“pathways to impact”. Kearnes and Wienroth interpret this shift in framing as an “… alternative
theorisation of the relationship between research and socio-economic impact in which basic science
is cast as ‘underpinning’ long-term social impacts and an attempt to generate new metrics that can
quantify the cumulative and non-linear effects of a broad portfolio of publicly resourced research”
(Kearnes and Wienroth 2011, p. 167).
The assessment of the dissemination of research can be seen as a type of formative or process
oriented evaluation, in contrast to summative or outcome oriented evaluation. It is important to
distinguish between these two types as they serve different purposes and require different
14
methodological approaches. In a strict sense, process oriented evaluations do not measure effects of
an intervention (Ramberg and Knall 2012).
The interaction between research and society has been the focus also for the FP7 SIAMPI project (see
Spaapen and Van Drooge 2011). The SIAMPI approach is based on the concept of productive
interactions as a way of assessing the social impact of research. The model assumes that for social
impact to take place there needs to be contact between researchers and non-academic stakeholders.
When this contact leads to an effort by the stakeholder to engage with research a productive
interaction is considered to have taken place. When the stakeholder does something new or in a
different way based on these productive interactions, research can be said to have had an impact.
Another similar approach is developed in an article on “public value mapping” by Bozeman and
Sarewitz.2 The aim is to “… provide an alternative to ‘market failure’ thinking that has been so
powerful in science policy-making” (Bozeman and Sarewitz 2011). They argue that to assess the
capacity of research programmes to achieve (non-economical) social goals we need to map public
values. Several methodologies can be used for this purpose, for example, the application of a set of
criteria that makes it possible to identify public values failure, i.e., “… when neither the market nor
public sector provides goods and services required to achieve public values” (Bozeman and Sarewitz
2011). The model has largely a case-based approach.
Expert reviews
One way of assessing the impact of research is to obtain information from groups of people that have
special insight into the field in question. This form is often used in combination with other methods.
The information can be gathered in different ways.
Expert panel
An expert panel with relevant experience can both contribute to estimating what difference research
has made and to giving feedback on how possible pathways have been used. As in the case with peer
review there are several methodological problems that have to be considered.
One specific method using experts that is interesting in impact assessments is the Delphi method. It
is a communication technique that uses experts, and the purpose is to get informed predictions.
Opinion collection is achieved by conducting a series of surveys using questionnaires. The result of
each survey will be presented to the group and the questionnaire used in the next round is built
upon the result of the previous round.
Anecdotes
Even if anecdotal evidence cannot be quantified it can help to identify some aspects of research
impact. Anecdotes that relate how a certain research project has benefited the society in some way
also illustrate the variety of possible impacts.
Surveys 2 The article is published in a special feature issue of Minerva where the other articles present examples of
implementation of “public value mapping”.
15
Stakeholder surveys
As stakeholders by definition represent certain interests, there are shortcomings with surveying to
determine impact. However, there are also positive features: “… stakeholders’ perceptions of impact
and the willingness of a university to work with outside agencies can be valuable … Such surveys can
also help identify ways of increasing impact by improving linkages, changing perceptions and
removing impediments to the flow, exchange and use of knowledge.” (Go8 2011, p. 15).
Commercialisation surveys
“… commercialisation surveys normally collect quantitative data relating to matters such as the
number of staff devoted to technology transfer, spin-off companies, patents and other IP rights
applied for or granted, and licensing income.” (Go8 2011, pp. 15-16). Although this kind of survey
provides quantitative data, it has to be critically analysed due to different types of biases. “Moreover,
commercialisation as measured by surveys provides a very narrow perspective on research impact
even in the confined context of achieving impact through business.” (Go8 2011. p. 16).
Case studies
Case study as a concept is defined and used in several ways. Looked upon as a method, case studies
in themselves often encompass a combination of different quantitative and qualitative methods.
What characterises case studies is the detailed analysis of individual research projects, programmes
or individual research institutes, etc. In this way case studies, like some other non-experimental
methods, allow for relevant contextual factors in contrast to RCT designs. Case studies have the
advantage of giving detailed information in a process oriented manner. That the level of aggregation
is low can, however, be a limitation of its usefulness. There is also much to gain especially as case
studies often provide us with insights that can help us to develop a better understanding of the
process of research impact. In other words a methodology well suited for formative or process
oriented evaluation.
A general challenge with case studies is the selection of cases. If the aim of the study is to use the
cases (e.g., research projects within a research programme) as representatives for a full research
programme, care has to be taken so that the cases really mirror the full programme. But case studies
can be used for other aims. For example, if the aim of the impact assessment is to investigate if
excellent scientific research has more or less impact than good scientific research, cases should be
selected that represent this. And if the aim is to get a better understanding of how research
transforms into impact (or not) it is appropriate to select cases that are known to or are expected to
generate impact.
In this context the relatively well-known “payback model” could be mentioned. The model is built
around case studies, and is a tool to facilitate data collection (surveys, interviews and document
studies) and provides a common structure for each case study, and thereby facilitates a cross-case
analysis. It was originally designed to capture socioeconomic impact of health services research, but
has been adapted and applied in a number of studies outside health and medical research. It consists
of two elements. Firstly, there is a model of the research process (from research idea to impact on
people and society) indicating when impacts can be expected. Secondly, there are categories of
benefits from research in which paybacks or impacts can be classified. They include both benefits
16
associated with the academic world (knowledge production and research capacity) and wider
benefits for society. These paybacks exist in somewhat different categories depending on the focus
of the impact assessment. In general terms they can be seen as (see Klautzer et al. 2011, compare
with Hanney et al. 2004 where the payback categories are directed towards the health sector):
a) Knowledge (explicit and codified knowledge)
b) Impact on future research (capacity building, new methods, career development)
c) Impacts on policy (impact on policy making at national level within professional
bodies and organisations)
d) Impacts on practice (individual behaviour)
e) Wider social and economic impacts.
Hindsight studies
Hindsight studies are a very special type in that they aim to trace the links backwards from the
identified impact to the research that contributed to the impact. Although hindsight studies can
appear to be less problematic than foresight studies, the critical points – attribution, time frame and
the counterfactual argument – are as relevant for these studies too.
Economic Models
One way to approach the interest in assessing value for money is different types of economic
models. Public policy evaluation from an economist’s viewpoint is concerned with the goals for public
policy and priority settings. The computing power available today and the amount of statistics
collected makes it possible to use complex systems modelling in studying the intricate ways in which
science has an impact on society. While a regression-based quasi-experimental method is highly
attractive for economists, agent-based and network-based models might provide a new and better
way to assess and guide research and innovation policies (Ramberg and Knall 2012, pp. 13-16).
Econometric analysis
Econometric analysis is primarily used for assessing impacts of research on a macro level. Advanced
numerical analysis techniques based on large amounts of data give information about the overall
economic impact that can help justify government funding of research. It does not, however, capture
the more intangible impacts of research and it is not applicable for assessing the performance of
individual research agencies or research programmes.
Cost-benefit analysis
Cost-benefit analysis is suitable to assess impact from projects and programmes, but can also be
used in a broader context. The objective is to compare the costs of investment in research to the
estimated economical benefits of the research. There are many methodological challenges with cost-
benefit analysis, such as the time frame and the attribution problem, and the quality of different
studies can vary a lot. There is also the question of what kind of assumptions have to be made in
order to define indicators and a model of causality. It is crucial that attention is paid to this kind of
methodological challenge as the assumptions that the model is based on will highly influence the
conclusions that can be drawn.
17
Questions asked and appropriate methods
Which purpose the impact study aims to fulfil affects the choice of method, and makes some
approaches more appropriate than others. For example, the payback model is more oriented
towards showing value for money. The SIAMPI approach is more directed towards the improvement
of understanding the process of how research transforms into impact. For example, in one study
using the SIAMPI model, one of the results was that researchers whose research was assessed
changed their future research practice taking the perspective of stakeholders into account and
stakeholders increased their contacts with researchers (Molas-Gallart and Tang 2011). Consequently,
it is difficult on a general level to make recommendations on which method to use. The table below
gives some indication of which method to use for a certain question, although it should not be seen
as an exhaustive list.
Relevant Questions Methods for Answering Questions
How much has been spent thus far? Does the progress achieved thus far match expectations based on those expenditures?
• Cost-benefit analysis
How are resources to be transformed into desired outputs and outcomes?
• Peer review/Expert judgment • Case study • Econometric studies
Is the programme’s research of high scientific quality? Is it relevant, productive and well managed?
• Peer review/Expert judgment
What relationships are developing? Is the programme strengthening the research network?
• Network analysis • Before-and-after applications
What additional project-related relationships have developed among researchers?
• Network analysis • Before-and-after applications
How are programme mechanisms, processes, and/or activities working? How can they be strengthened?
• Monitoring activities • Case study - descriptive/exploratory • Econometric studies
What are the programme’s codified knowledge outputs?
• Bibliometrics
How does the programme’s output productivity compare with similar programmes?
• Benchmarking
How noteworthy are the resulting patents? What are the hot trends? Are there important regional impacts?
• Hot-spot patent analysis
To what extent have the programme’s outputs • Indicators
18
been commercialised? • Technology commercialisation tracking
What factors are influencing industry’s adoption/lack of adoption of the programme’s technologies?
• Case study - descriptive/explanatory
What are the realised benefits and costs of the technology to date? What share of net benefits from the technology is attributed to the programme?
• Benefit-cost analysis
How is the programme working thus far? • Case study - descriptive/explanatory
Are there one or more noteworthy consequences that can be shown to link back directly to the research?
• Hindsight studies, Historical tracing (including citation analysis)
If we had it to do all over again, would we have launched the programme or initiative?
Peer review/Expert judgment supported by multiple retrospective evaluation methods
What benefits are there for society in general? Is the action (e.g., a research programme) showing value for money?
Economic models (on an aggregate level)
Case studies (based on research projects, for example, the payback model)
How does research affect policy makers? The SIAMPI approach
What health gains does research create? The payback framework
How is the knowledge produced used in policy? Interviews, document analysis
5. Recommendations
It is clear to see that impact is difficult to assess, and there are several ways to conduct an impact
evaluation. There is a wide discussion about “best practice” or “good practices”. In this chapter the
working group will summarise the recommendations on when and how to perform impact studies. It
should be noted that these recommendations are primarily written for research funders and
research performing organisations.
Assessments of impact are an important element in demonstrating, understanding and facilitating
the value of research. Furthermore, impact should be seen as a broad concept, both in regard to in
what areas impact can occur and what is to be considered as impact. This was discussed in detail in
chapter 2.
19
However, an assessment of impact faces several methodological challenges and we have above – on
a general level - dealt with how to cope with the most demanding ones. On a more specific level –
when a certain assessment of impact is to be performed – there is no single method to be preferred.
Instead it is a matter of finding an appropriate method, taken the recommendations on the general
level into account, depending on what kinds of impacts are assessed. Furthermore, it is also,
considering the difficulty and complexity to assess impact, advisable not to stick to one method in an
impact assessment but to use different approaches.
It is also advisable to see the potential in impact evaluations as a tool to create better informed
policies by research funders, better opportunities for the creation of scientific knowledge and the
facilitation of the process where scientific knowledge produces benefits for the wider community.
This does not mean that impact evaluations are an appropriate tool for showing value for money, but
it is advisable that impact evaluations are constructed so that they have something to say about how
benefits for society are better achieved. Another conclusion and recommendation is that the first
thing to do when conducting an impact assessment is to carefully specify and select what is to be
assessed rather than how to assess. This is not to say that how to measure is not important, it is very
important. But it plays second fiddle to the question of why and what to measure.
A current trend in impact assessments is to find indicators of future impact. This is an area that is
fruitful to develop further, because if there are stable indicators that serve as a proxy for longer-term
impact assessments will be an even more useful instrument to monitor programmes and make
better informed policies. However, as argued earlier, it is still important to continue to do studies
that employ a longer perspective on how research transforms into impact.
There is also the question of cost for the impact evaluations. Obviously, the cost should never exceed
the benefits that can be attributed to the impact study. It may be difficult to estimate the benefits,
but this perspective must be taken into account. Also, the costs of the activities being assessed must
be put in proportion to the cost of the study. Consequently, as in all forms of evaluations, before an
impact study is carried out its usefulness must be put in perspective to its cost and the cost of what is
being assessed.
There are also some further recommendations that the working group would like to highlight:
Continuing cooperation between funding agencies and researchers on impact evaluation, as
this will strengthen the activities both of research funders and researchers.
Funding agencies should prepare for impact evaluations that are long term and also include a
wide range of different impacts. A monitoring of impacts will provide valuable knowledge to
further impact assessments and to the process of how research turns into gains to society.
More focus on how to establish the counterfactual position, as this is one of the areas that
can be stronger in impact assessments of research.
Perform an impact evaluation that involves different countries and research agencies. It
could be similar programmes that are assessed with the same approach. This could be the
next step for European cooperation within the area of research evaluations.
20
References
Academy of Finland (2009). Impact Evaluation of Finnish Programmes for Centres of Excellence in
Research 2000–2005 and 2002–2007.
Bozeman, B. and Sarewitz, D. (2011). Public Value Mapping and Science Policy Evaluation, Minerva
49, pp. 1-23.
Canadian Academy of Health Sciences (2009). Making an Impact – A Preferred Framework and
Indicators to Measure Returns on Investment in Health Research.
Davies, H., Nutley, S. and Walter, I. (2005). Assessing the impact of social science research – conceptual, methodological and practical issues. A background discussion paper for ESRC Symposium on Assessing Non-Academic Impact of Research, May 2005, 24 p.
Delanghe, H. and Teirlinck, P. (2010). Optimizing the Policy Mix by the Development of a Common
Methodology for the Assessment of (Socio-) Economic Impacts of RTDI Public Funding, Idea Consult.
Donovan, C. (2007a). Introduction: Future pathways for science policy and research assessment:
metrics vs peer review, quality vs impact, Science and Public Policy, 34(8), pp. 538-542.
Donovan, C. (2007b). The qualitative future of research evaluation, Science and Public Policy, 34(8),
pp. 585-597.
European Commission (2010). A vision for strengthening world-class research infrastructures in the
ERA.
European Science Foundation (2009). Evaluation in National Research Funding Agencies: approaches,
experiences and case studies - A report of the ESF Member Organisation Forum on Ex-Post Evaluation
of Funding Schemes and Research Programmes.
Exceptional Returns: The value of Investing Health R&D in Australia II (2008). Prepared for the
Australian Society for Medical Research by Access Economic, Canberra 2008.
Go8 (2011). Measuring the impact of research – the context for metric development, Go8
Backgrounder 23, 20 p. Australia, The Group of Eight. www.go8.edu.au
Grant, J. et al. (2009). Documented Briefing, Capturing Research Impacts - A review of international practice. RAND Europe: DB-578-HEFCE Prepared for the Higher Education Funding Council for England. http://www.hefce.ac.uk/media/hefce/content/pubs/2009/rd2309/rd23_09.pdf Hanney S.R., Grant, J., Wooding, S. and Buxton M.J. (2004). Proposed method for reviewing the
outcomes of health research: the impact of funding by the UK’s Arthrtis Research Campaign, Health
Research Policy and Systems, 2004, 2:4.
21
Kearnes, M. and Wienroth, M. (2011). Tools of the Trade: UK Research Intermediaries and the Politics
of Impacts, Minerva 49, pp. 153-174.
Klautzer, L., Hanney, S., Nason, E., Rubin, J., Grant, J. and Wooding, S. (2011.) Assessing policy and
practice impacts of social science research: the application of the Payback Framework to assess the
Future of Work Programme”, Research Evaluation 20 (3).
Levitt, R. et al. (2010). Assessing the impact of arts and humanities research at the University of
Cambridge, RAND Europe.
Langfeldt, L., Brorstad Borlaug, S. and Gulbrandsen, M. (2010). The Norwegian Centre of Excellence
Scheme – Evaluation of Added Value and Financial Aspects, Rapport 29/2010. NIFU STEP, 2010.
Luoma, P. et al. (2011). Better results, more value – A framework for analysing the societal impact of
Research and Innovation. A Tekes Review 288/220.
Medical Research: What’s it worth? Estimating the economic benefits from medical research in the
UK (2008). Health Economics Research Group (HERG), Brunel University, Office of Health Economics
(OHE), RAND Europe.
Molas-Gallart, J., Tang, P. and Morrow, S. (2000). Assessing the non-academic impact of grant-funded
socio-economic research: results from a pilot study, Research Evaluation 9(3), pp. 171–182.
Molas-Gallart, J. and Tang, P. (2011). Tracing ’productive interactions‘ to identify social impacts: an
example from the social sciences, Research Evaluation 20(3), pp. 219–226.
Ramberg, I. and Knall, M. (2012). Challenges measuring effects of research and innovation policy
interventions in ex-post impact evaluations. A synthesis report. 4 March 2012. Oslo: NIFU.
RAND Europe (2004). The returns from arthritis research - A report prepared by RAND Europe for the
Arthritis Research Campaign (arc).
REF Pilot a (2009). Identification and dissemination of lessons learned by institutions participating in
the Research Excellence Framework (REF) bibliometrics pilot. Results of the Round Two Consultation –
report to HEFCE by Technopolis. 2009.
http://www.hefce.ac.uk/media/hefce/content/pubs/2009/rd1809/rd18_09.pdf
REF Pilot b (2010). REF Research Impact Pilot Exercise Lessons-Learned Project: Feedback on Pilot
Submissions. Final report. www.technopolis-group.com. 2010.
http://www.ref.ac.uk/media/ref/content/pub/refresearchimpactpilotexerciselessons-
learnedprojectfeedbackonpilotsubmissions/re02_10.pdf
REF 2014 c (2010). Research Excellence Framework impact pilot exercise: Findings of the expert
panels. A report to the UK higher education funding bodies by the chairs of the impact pilot panels.
2010.
http://www.ref.ac.uk/media/ref/content/pub/researchexcellenceframeworkimpactpilotexercisefindi
ngsoftheexpertpanels/re01_10.pdf
22
Spaapen, J. and Van Drogge, L. (2011). Productive interactions as a tool for social impact assessments
of research, Research Evaluation, 20(3).
Swedish Research Council (2010). Evaluation of Impacts of Medical Research, Stockholm, Swedish
Research Council: Report Series 9:2010.
Tassy, G. (2003). Methods for Assessing the Economic Impacts of Government R&D, USA, National
Institute of Standards & Technology: Planning Report 03-1.
UK Evaluation Forum (2006). Medical Research: assessing the benefits to society.
List of MO Forum Working Group Members
List of Members of the WG
Chair: Per Janson Swedish Research Council (VR)
Members: Mustafa Ay The Scientific and Technological Research Council of Turkey
(TÜBITAK)
Frank Bingen Fonds National de la Recherche (FNR), Luxembourg
Sarah Chen The National Research Council, Italy (CNR)
Helen O’Connor &
Graham Love
Science Foundation Ireland (SFI)
Christian Fischer Austrian Science Fund (FWF)
Marianne Gauffriau Danish National Research Foundation (DNRF)
Isabelle Henry Institut national de la Santé et de la Recherche Médicale
(Inserm)
Inger Jonsson Swedish Council for Working Life and Social Research (FAS)
Paul reeves BBSRC
Jaana Roos Academy of Finland
Mike Mugabushaka European Research Council, Observer
Briony Rayfield &
Liz Allen
The Wellcome Trust, Observers
23
Acknowledgements
This report has been compiled by Per Janson (VR). ESF is grateful to him – the Co-Chair of the MO
Forum Working Group - and to the significant input from working group members who contributed
to lively and informative discussion in the ESF Member Organisation Forum over the last four years
as well as for its management to Laura Marin (ESF). In particular, the drafting of this document relied
upon the contributions of Sarah Chen (CNR), Inger Jonsson (FAS), Mustafa Ay (Tübitak), Helen
O’Connor (SFI), Marianne Gauffriau (DNRF) and Isabelle Henry (Inserm).